A concrete dam high-precision deformation monitoring method and system fusing GNSS and external environment parameters

Abstract

The application provides a concrete dam high-precision deformation monitoring method and system fusing GNSS and external environment parameters, relates to the dam deformation monitoring technical field, and obtains first displacement data and environment data of a dam monitoring point, and generates a dam surface model in combination with photogrammetry data. Original displacement data is obtained by measuring based on the surface model, a compensation model is used and environment data is fused, and real deformation data of the dam is calculated. The real deformation data is combined with externally obtained ground surface deformation data and settlement data, comprehensive deformation data and settlement distribution data which can comprehensively reflect the dam state are generated. Key features are extracted from the comprehensive data, the features are processed, abnormal patterns are recognized, and future deformation trends are predicted. According to the recognized abnormalities and predicted trends, corresponding early warning information is generated. The application can realize high-precision, multi-source fusion monitoring and trend early warning of dam deformation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of dam deformation monitoring, and in particular to a concrete dam high-precision deformation monitoring method and system fusing GNSS and external environmental parameters. BACKGROUND

[0002] As a key water conservancy infrastructure, the structural safety of a concrete dam is of great importance. Therefore, high-precision deformation monitoring of the dam is a core means to ensure its long-term stable operation and prevent catastrophic accidents.

[0003] At present, dam deformation monitoring mainly relies on traditional geodetic techniques such as global navigation satellite system (GNSS), precise leveling or total station. These methods usually obtain three-dimensional coordinate changes of discrete points by monitoring points arranged on the dam surface to reflect the displacement of the dam.

[0004] However, the displacement data obtained by these monitoring methods contains non-structural deformation components caused by external environmental changes such as temperature and water level. The lack of data fusion makes it difficult to accurately separate the real structural deformation of the dam, thereby affecting the accuracy of the dam safety state assessment. Therefore, there is a technical problem in the prior art that the monitoring results cannot comprehensively and accurately reflect the real deformation state of the dam. SUMMARY

[0005] The purpose of the present application is to provide a concrete dam high-precision deformation monitoring method and system fusing GNSS and external environmental parameters to solve the technical problem in the prior art that the monitoring results cannot comprehensively and accurately reflect the real deformation state of the dam.

[0006] To solve the above technical problems, in a first aspect, the present application provides a concrete dam high-precision deformation monitoring method fusing GNSS and external environmental parameters, comprising:

[0007] obtaining first displacement data and environmental data, the first displacement data being displacement data of dam monitoring points collected by a GNSS receiver, and the environmental data including temperature data, humidity data and water level data;

[0008] obtaining photogrammetry data, generating a dam surface model based on the photogrammetry data and the first displacement data;

[0009] based on the dam surface model, performing measurement to obtain original displacement data, and based on the original displacement data and the environmental data, using a compensation model to generate second displacement data; the second displacement data being deformation data of the dam;

[0010] obtain ground surface deformation data and settlement data, generate comprehensive deformation data and settlement distribution data based on the ground surface deformation data, the second displacement data and the settlement data;

[0011] extract key features from the comprehensive deformation data and the settlement distribution data, process the key features to identify abnormal patterns and predict deformation trends;

[0012] generate early warning information based on the abnormal patterns and the deformation trends.

[0013] Optionally, the generating comprehensive deformation data and settlement distribution data based on the ground surface deformation data, the second displacement data and the settlement data comprises:

[0014] spatially superimpose the second displacement data and deformation amount data of multiple points in the ground surface deformation data to generate comprehensive deformation data;

[0015] generate a continuous deformation surface based on the comprehensive deformation data and settlement data of multiple points in the dam and foundation body, and using a scattered point-based surface reconstruction method;

[0016] extract settlement contour lines from the continuous deformation surface to generate settlement distribution data.

[0017] Optionally, the generating a continuous deformation surface based on the comprehensive deformation data and settlement data of multiple points in the dam and foundation body, and using a scattered point-based surface reconstruction method comprises:

[0018] combining the comprehensive deformation data and the settlement data of multiple points in the dam and foundation body into a scattered point data set;

[0019] processing the scattered point data set to generate the continuous deformation surface using a moving least squares method, the processing process comprising: for any target point on the continuous deformation surface, locally fitting a surface to the scattered point data set in a neighborhood of the target point, and determining a value of the target point based on a result of the local surface fitting;

[0020] wherein the neighborhood is determined by a search radius parameter, and the search radius parameter is adjusted according to the density and distribution characteristics of the monitoring points in the dam foundation and dam abutment area.

[0021] Optionally, the extracting settlement contour lines from the continuous deformation surface to generate settlement distribution data comprises:

[0022] extracting settlement contour lines from the continuous deformation surface according to a set settlement amount interval;

[0023] The extracted settlement contour is smoothed, and adjacent contour segments are topologically connected to generate a settlement contour map as settlement distribution data.

[0024] Optionally, the key features are extracted from the comprehensive deformation data and the settlement distribution data, and the key features are processed to identify abnormal patterns and predict deformation trends, including:

[0025] Key features are extracted from the comprehensive deformation data and the settlement distribution data; the key features include node attributes and topological connection relationships; the node attributes include spatial coordinates, settlement values, and deformation values; and the topological connection relationships are used to represent the spatial distribution of the settlement distribution data;

[0026] The key features are input into an autoencoder network, the key features are mapped to a low-dimensional latent space through an encoder in the autoencoder network, and reconstructed features are output from the low-dimensional latent space through a decoder in the autoencoder network;

[0027] Based on the mean square error, the reconstruction error between the key features and the reconstructed features is calculated;

[0028] Based on the reconstruction error and a dynamic threshold, the abnormal patterns are identified; the dynamic threshold is set based on historical normal data;

[0029] Based on historical deformation data, the deformation trends are predicted.

[0030] Optionally, based on the dam surface model, measurements are taken to obtain original displacement data, and based on the original displacement data and the environmental data, a second displacement data is generated using a compensation model, including:

[0031] Based on the dam surface model, the measurement point positions of the total station are determined;

[0032] The total station is driven to perform laser measurements on the measurement point positions, and through multiple laser measurements, a three-dimensional coordinate sequence of each measurement point position is obtained to generate a displacement change sequence, and the displacement change sequence is combined into original displacement data;

[0033] A compensation model of environmental parameters and displacement changes is established, and the compensation model is established based on a multiple regression model;

[0034] The compensation model is used, and based on the environmental data, the original displacement data is adjusted to generate real deformation information of the dam, and the real deformation information of the dam is taken as second displacement data.

[0035] Optionally, the using the compensation model and adjusting the original displacement data based on the environment data to generate the real deformation information of the dam comprises:

[0036] The compensation model is used to calculate the environment displacement component with temperature, humidity and water level change as independent variables and with the environment displacement component as dependent variable.

[0037] The environment displacement component is subtracted from the original displacement data to generate the real deformation information of the dam.

[0038] Optionally, the acquiring photogrammetry data and generating the dam surface model based on the photogrammetry data and the first displacement data comprises:

[0039] The dam surface is imaged from multiple angles by a camera carried by a UAV to acquire an image data set.

[0040] Feature points are extracted from the image data set, and an initial three-dimensional point cloud is calculated according to the matching relationship of the feature points.

[0041] The first displacement data is used as a control point to correct the position of the initial three-dimensional point cloud to generate a corrected three-dimensional point cloud.

[0042] Based on the corrected three-dimensional point cloud, a dense point cloud is generated, and based on the dense point cloud, a dam surface model is constructed.

[0043] In a third aspect, the application provides a concrete dam high-precision deformation monitoring system fusing GNSS and external environment parameters, comprising:

[0044] An acquisition module is configured to acquire first displacement data and environment data, wherein the first displacement data is displacement data of dam monitoring points, and the environment data comprises temperature data, humidity data and water level data.

[0045] A generation module is configured to acquire photogrammetry data, generate a dam surface model based on the photogrammetry data and the first displacement data, measure based on the dam surface model to obtain original displacement data, and generate second displacement data based on the original displacement data and the environment data and using a compensation model, wherein the second displacement data is deformation data of the dam, acquire ground surface deformation data and settlement data, and generate comprehensive deformation data and settlement distribution data based on the ground surface deformation data, the second displacement data and the settlement data.

[0046] An identification module is configured to extract key features from the comprehensive deformation data and the settlement distribution data, process the key features to identify abnormal patterns and predict deformation trends.

[0047] The generation module is further configured to generate early warning information based on the abnormal pattern and the deformation trend.

[0048] In a third aspect, the present application provides an electronic device, comprising:

[0049] a memory configured to store a computer program;

[0050] a processor configured to implement the steps of the method for high-precision deformation monitoring of a concrete dam by fusing GNSS and external environmental parameters when executing the computer program.

[0051] The method for high-precision deformation monitoring of a concrete dam by fusing GNSS and external environmental parameters provided by the present application can realize the synchronous collection of dam foundation displacement and external environmental influencing factors by obtaining first displacement data and environmental data of the dam monitoring points; can construct a high-precision digital twin dam by obtaining photogrammetry data and combining the first displacement data to generate a dam surface model, thereby providing an accurate three-dimensional reference for subsequent automated measurement; can accurately separate the real deformation caused by structural reasons by measuring based on the surface model and using a compensation model to generate second displacement data, thereby excluding the interference of environmental factors; can integrate the dam body deformation and the changes of the surrounding geological environment to form a comprehensive understanding of the overall deformation field by obtaining ground deformation data and settlement data and fusing them with the second displacement data; can mine deep-level deformation rules from massive multi-source data by extracting key features from the comprehensive data and processing them, thereby realizing intelligent diagnosis of the deformation state and scientific prediction of future trends; and can provide a direct basis for the safety management and maintenance decision of the dam by generating early warning information based on the identified abnormalities and predicted trends.

[0052] Further, by limiting the specific fusion processing method of multi-source data, the accurate construction method of the compensation model, and the intelligent algorithm of feature processing, it is ensured that each step from data collection, model compensation to feature analysis has high precision and high reliability, thereby ultimately improving the automation level of the entire monitoring method and the accuracy of the early warning results. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0054] Figure 1A flowchart of a high-precision concrete dam deformation monitoring method fusing GNSS and external environmental parameters is provided for the embodiments of the present application.

[0055] Figure 2 A flowchart of a specific embodiment of a high-precision concrete dam deformation monitoring method fusing GNSS and external environmental parameters is provided for the embodiments of the present application.

[0056] Figure 3 A structural diagram of a high-precision concrete dam deformation monitoring system fusing GNSS and external environmental parameters is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0057] Existing dam deformation monitoring relies on a single or small number of measurement methods (such as GNSS, total station, and leveling), and the data coverage is limited. Moreover, the displacement data is mixed with environmental influences such as temperature, humidity, and water level changes, making it difficult to distinguish between "structural real deformation" and "environmentally induced deformation". In addition, the monitoring data sources are scattered, the caliber is different, the cross-source fusion is insufficient, and the evaluation results are unstable, the trend judgment is unreliable, and it is difficult to support timely and accurate safety warning.

[0058] The present application proposes a high-precision monitoring method fusing GNSS and external environmental parameters: synchronously collecting displacement and environmental data, combining photogrammetry to generate a dam surface model; conducting measurement on the surface model and introducing a compensation model to remove environmental influences to obtain real deformation data; further fusing with ground deformation data and settlement data to form comprehensive deformation data and settlement distribution data; finally extracting key features, identifying abnormalities, and predicting trends to output warning information. This scheme significantly improves spatial coverage, data consistency, and result interpretability through "multi-source data fusion + environmental compensation", effectively solving the problems of real deformation being difficult to separate, monitoring results being not accurate enough, and early warning being lagging in the prior art.

[0059] In order for those skilled in the art to better understand the present application, the present application will be further described in detail below in combination with the drawings and specific embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0060] The core of the present application is to provide a high-precision concrete dam deformation monitoring method fusing GNSS and external environmental parameters, and a flowchart of a specific embodiment of the method is shown in Figure 1 The method comprises the following steps.

[0061] S101, acquiring first displacement data and environmental data.

[0062] The GNSS receiver is a device for receiving global navigation satellite system signals to determine a position. The first displacement data is displacement data of a dam monitoring point collected by the GNSS receiver, and specifically can include a sequence of position changes of the monitoring point in three directions in space. The environmental sensor is a device for collecting the state of the surrounding environment, including a temperature sensor, a humidity sensor, and a water level sensor, and the environmental data is the temperature, humidity, and water level data collected by these sensors.

[0063] In the embodiments of the present application, first, GNSS receivers are arranged at multiple monitoring points of the dam, each GNSS receiver continuously receives satellite signals and calculates the three-dimensional coordinates of the point, and these time-sequentially arranged three-dimensional coordinates constitute the displacement data. For example, one GNSS receiver is arranged at each of the dam crest, dam waist, and other positions prone to deformation, and the three-dimensional coordinates of the position are calculated every 10 seconds, so that a large amount of three-dimensional coordinate data of the point position arranged in chronological order can be obtained in one day.

[0064] Secondly, environmental sensors are arranged at the above monitoring points or nearby positions, wherein the temperature sensor and the humidity sensor are installed on the surface of the dam body close to the GNSS receiver, and the water level sensor is installed at the water level monitoring point near the dam. These sensors work synchronously with the GNSS receiver and continuously collect temperature, humidity, and water level data as environmental data. For example, the temperature sensor collects environmental temperature every 10 seconds, and the humidity sensor and the water level sensor also collect corresponding data at the same frequency.

[0065] Finally, the time stamp of each data in the three-dimensional coordinate sequence (i.e. the original information constituting the first displacement data) obtained by each GNSS receiver is matched with the time stamp of the temperature, humidity, and water level data (i.e. the environmental data) collected by the corresponding point environmental sensor, so that the three-dimensional coordinate data and the environmental data collected at the same time are associated, and preparation is made for subsequent data analysis and processing. For example, the three-dimensional coordinates collected at 10:00:00 at a monitoring point are bound together with the temperature, humidity, and water level data collected at the same time at the point.

[0066] S102, acquiring photogrammetry data, and generating a dam surface model based on the photogrammetry data and the first displacement data.

[0067] Optionally, step S102 can specifically include the following steps:

[0068] S1021, acquiring image data sets by a camera carried by a UAV for multi-angle image acquisition of the dam surface.

[0069] The photogrammetry data refers to dam surface image data photographed by a camera carried by the unmanned aerial vehicle. The image data set is all image files collected after the unmanned aerial vehicle photographs from different angles and can cover the whole dam.

[0070] S1022, extracting feature points from the image data set, and calculating an initial three-dimensional point cloud according to a matching relationship of the feature points.

[0071] The feature points are points with obvious recognition in the image, such as crack corners, convex edges and the like on the dam surface. The initial three-dimensional point cloud is preliminary three-dimensional shape data of the dam surface composed of a large number of three-dimensional coordinate points calculated according to the matching relationship of the feature points.

[0072] S1023, using the first displacement data as a control point to correct the position of the initial three-dimensional point cloud to generate a corrected three-dimensional point cloud.

[0073] The control point is a three-dimensional coordinate point with accurate position selected from the first displacement data and used to correct the position of the initial three-dimensional point cloud. The corrected three-dimensional point cloud is three-dimensional point cloud data generated after the position of the initial three-dimensional point cloud is corrected by using the control point.

[0074] S1024, generating a dense point cloud based on the corrected three-dimensional point cloud, and constructing a dam surface grid model based on the dense point cloud.

[0075] The dense point cloud is a more dense three-dimensional coordinate point set formed by supplementing more details on the basis of the corrected three-dimensional point cloud. The dam surface model is a model that can intuitively show the shape of the dam surface formed by connecting the dense point cloud into a grid according to a certain rule.

[0076] In the embodiment of the application, first, the flight route of the unmanned aerial vehicle is planned according to the length, height and shape of the dam in step S1021 to ensure that the route can cover every area of the dam, and then the unmanned aerial vehicle carrying the camera is controlled to fly according to the planned route. The camera continuously photographs according to the set photographing interval and angle during the flight process to obtain the image data set that can completely cover each area of the dam. For example, for a dam with a height of 100 meters and a length of 500 meters, three horizontal routes (at 20 meters, 60 meters and 100 meters of dam height respectively) and two longitudinal routes (at 50 meters on the left and right sides of the dam body respectively) are planned, the camera is set to photograph one picture every 2 seconds, and after the unmanned aerial vehicle flies along the route, all the photographed images are collected to form the image data set.

[0077] Secondly, by step S1022, the feature extraction technology is applied to process each image in the image dataset, and points with obvious features such as corners of dam surface and hole edges are found as feature points. Then, the feature matching algorithm is used to pair feature points representing the same position in different images. According to the two-dimensional coordinates of the paired feature points in different images and the parameters when the camera is shooting, the three-dimensional coordinates of each feature point are calculated. The three-dimensional coordinates of all feature points are summarized to form an initial three-dimensional point cloud. For example, the corner point of the inspection platform on the dam surface is extracted as a feature point from the image, the corner point in different angle images is matched, and the three-dimensional coordinates of the corner point are calculated by combining the focal length, shooting angle and other parameters of the camera. A large number of such points constitute the initial three-dimensional point cloud.

[0078] In a specific embodiment, a feature detection algorithm such as SIFT, SURF or ORB can be used to automatically identify and extract thousands of stable and easily distinguishable feature points in images taken from different angles. Then, through feature description and matching algorithm, the system finds a pair of "same-name feature points" corresponding to the same physical space point in two or more different images. For example, the end point of a crack on the dam surface is captured at the (X1, Y1) pixel position in the first photo and at the (X2, Y2) pixel position in the second photo. These two points constitute a pair of same-name feature points.

[0079] The next step is the core calculation link of triangulation. The basic geometric principle can be understood as follows: two observers (two cameras) observe the same object (feature point) from different positions at the same time. By connecting the baseline between the two observer positions and the line of sight of each observer to the object, a triangle can be formed. As long as the precise positions of the two observers, the distance between them (baseline), and their respective observation directions are known, the shape of the triangle in space can be uniquely determined, and the three-dimensional coordinates of the object can be calculated. In photogrammetry, the "line of sight" is the projection ray from the camera optical center to the pixel position of the feature point on the image.

[0080] Mathematically, this process is achieved by solving a system of equations. For a three-dimensional point P=(X, Y, Z) in space and a camera, the projection relationship between them can be described by the following camera model formula:

[0081]

[0082] wherein, is the two-dimensional homogeneous coordinates of the feature point on the image. is the homogeneous coordinates of the point in three-dimensional space. K is a 3x3 camera intrinsic parameter matrix containing focal length and principal point information, which is obtained through camera calibration. is a 3x4 camera extrinsic matrix, R is a rotation matrix, t is a translation vector, describing the pose and position of the camera in the world coordinate system, also solved by calibration or Structure from Motion in the processing. Z is a scale factor, representing the depth of the point.

[0083] When the same three-dimensional point P is observed in at least two images (camera 1 and camera 2), a system of equations can be established:

[0084]

[0085] In this system of equations, in addition to the coordinates (X, Y, Z) of the three-dimensional point P', all other parameters (R, t, z) are known. The optimal solution of the unknown , , , , , ) can be calculated by solving this overdetermined equation system by algebraic method (such as Direct Linear Transformation DLT or Singular Value Decomposition SVD). The optimal solution of the unknown is the coordinates (X, Y, Z) of the feature point in three-dimensional space. Repeat this process for all matched feature point pairs to obtain a large set of three-dimensional points, which are collectively referred to as the initial three-dimensional point cloud.

[0086] In practical applications, in order to improve accuracy, more than two images are usually used for joint calculation, and a global optimization algorithm (such as Bundle Adjustment) is used to iteratively optimize all three-dimensional point coordinates and camera parameters to minimize the "reprojection error" of all three-dimensional points in all images, thereby obtaining a more accurate and consistent initial three-dimensional point cloud.

[0087] Further, from the first displacement data of the monitoring points obtained in step S101, select the three-dimensional coordinates of the monitoring points distributed in different regions (such as the dam crest, dam waist, and dam foundation) of the dam as control points, and find the feature points corresponding to these control points in the initial three-dimensional point cloud. Compare the coordinates of the corresponding points in the initial three-dimensional point cloud with the control point coordinates, calculate the positional deviation between the two, and adjust the entire initial three-dimensional point cloud by translation, rotation, etc. to generate a more accurate corrected three-dimensional point cloud. For example, select the four control points of the dam crest midpoint, the dam waist left and right endpoints, and the dam foundation midpoint, and find that the initial three-dimensional point cloud midpoint is 3 meters ahead of the control point coordinates. The entire initial three-dimensional point cloud is translated 3 meters backward to complete the correction.

[0088] Finally, based on the corrected three-dimensional point cloud, a multi-view image matching method is used to compare the fine textures, concave-convex details, etc. on the dam surface in the images obtained in step S1021, calculate the three-dimensional coordinates of these details, add these new three-dimensional coordinate points to the corrected three-dimensional point cloud, form a dense point cloud containing more surface details, and then use a mesh generation technique to connect the points in the dense point cloud into multiple small triangular meshes according to the adjacent relationship, and these meshes are combined to build a dam surface mesh model.

[0089] In actual applications, for example, in the monitoring of a concrete dam, a camera is carried by a UAV, a flight plan containing 5 flight lines is planned, the dam is photographed from above the upstream, downstream, left abutment, right abutment and dam crest of the dam respectively, the flight height of each flight line is adjusted according to the height of the dam body, and a total of 600 images with a resolution of 4000x3000 are obtained to form an image dataset; the construction joint corners and ladder joints on the dam surface are extracted from the image dataset as feature points, the feature points in different images are matched through a feature matching algorithm, and initial three-dimensional point clouds are calculated; the three-dimensional coordinates of the left end point of the dam crest, the right end point of the dam crest, the middle point of the dam waist, the left end point of the dam foundation and the right end point of the dam foundation are selected as control points from the first displacement data of the 10 monitoring points obtained in step S101, and the initial three-dimensional point clouds are positionally corrected; based on the corrected three-dimensional point clouds, a multi-view image matching method is used to supplement the texture details of the dam body surface to form a dense point cloud, and a triangular mesh is connected from the dense point cloud using a mesh generation software to build a dam surface mesh model that can clearly show the shape and details of the dam body surface.

[0090] The overall scheme of S102 described above generates a dam surface model with accurate position and rich surface details through photogrammetry combined with first displacement data, not only intuitively presents the overall shape of the dam, but also provides accurate shape basis for subsequent driving of a total station for laser measurement, avoiding subsequent measurement deviation caused by inaccurate dam shape data.

[0091] S103, based on the dam surface model, measuring to obtain original displacement data, and based on the original displacement data and the environment data, using a compensation model to generate second displacement data.

[0092] Among them, the second displacement data is the deformation data of the dam.

[0093] Optionally, step S103 can specifically include the following steps:

[0094] S1031, based on the dam surface model, determining the measurement point of the total station.

[0095] The total station is a measuring instrument integrating light, machine, and electricity, and is used for measuring three-dimensional coordinates of point positions. The point positions are positions of the dam body selected based on a dam surface model and needing to be monitored.

[0096] S1032, driving the total station to perform laser measurement on the measurement point positions, obtaining a three-dimensional coordinate sequence of each measurement point position through multiple laser measurements, generating a displacement change sequence, and combining the displacement change sequence into original displacement data.

[0097] The displacement change sequence is a data sequence formed by arranging three-dimensional coordinates obtained through multiple measurements of the same measurement point position in time. The original displacement data is a set of original three-dimensional coordinate points combined by displacement change sequences of multiple measurement point positions and not compensated.

[0098] S1033, establishing a compensation model of environmental parameters and displacement changes, the compensation model being established based on a multiple regression model.

[0099] The compensation model is a mathematical model for establishing a quantitative relationship between environmental parameters and displacement changes. The second displacement data is data reflecting the deformation of the dam structure after eliminating the influence of environmental factors, i.e., the deformation data of the dam.

[0100] The training process of the compensation model first needs to construct a sample data set for model fitting. Specifically, the original displacement data measured by the total station and the environmental data such as temperature, humidity, and water level collected synchronously are accurately aligned through time stamps. Based on the data set, the change amount of the environmental data (such as temperature change amount , , water level change amount ) is used as the independent variable of the model, and the component of the total displacement caused by environmental factors is used as the dependent variable, and then a regression analysis technique such as least squares method is used for model fitting. This process will automatically calculate and solve a set of optimal model coefficients (for example , , , ), which quantitatively reveals the specific influence of the change of each environmental parameter on the displacement, thereby finally establishing a compensation model that can be used for subsequent displacement compensation calculation.

[0101] S1034, using the compensation model and based on the environmental data, adjusting the original displacement data to generate real deformation information of the dam, and taking the real deformation information of the dam as the second displacement data.

[0102] The step S1034 specifically comprises the following process: using the compensation model to calculate the environmental displacement component with temperature, humidity and water level change as independent variables and with the environmental displacement component as dependent variable; and subtracting the environmental displacement component from the original displacement data to generate real deformation information of the dam.

[0103] In the embodiments of the present application, first, the key regions of the dam body prone to deformation (such as the dam waist and dam foundation joint) are selected as the measurement points according to the dam surface model generated in step S102 in step S1031, and the coordinate information of these points is input into the total station control system to drive the total station to automatically align each measurement point for laser aiming.

[0104] Secondly, the measurement frequency and number of times of the total station are set in step S1032, and the three-dimensional coordinate sequence of each measurement point is obtained through multiple laser measurements to form a displacement change sequence. Then, the displacement change sequence data of all measurement points are collected, and these three-dimensional coordinate points are arranged according to their corresponding dam body positions to form the original displacement data. For example, a measurement point is measured continuously for 30 times to obtain a displacement change sequence containing 30 coordinates; and the sequence data of all points are integrated to form the original displacement data covering the key monitoring region of the dam.

[0105] Then, the environmental data (temperature, humidity, water level) collected in step S101 and the displacement change in the original displacement data generated in step S1032 are analyzed in association to establish a compensation model in step S1033. The specific process is as follows:

[0106] First, the displacement change sequence of each measurement point is aligned with the temperature, humidity and water level data collected synchronously according to the time stamp to form an environmental displacement data set containing independent variables (temperature T, humidity H, water level L) and dependent variables (displacement D).

[0107] Subsequently, based on the multiple regression model, an independent compensation model is established for each measurement point. The mathematical expression of the model is as follows: wherein: represents the environmental displacement component caused by the change of environmental parameters, , , represents the temperature, humidity and water level change relative to the reference value respectively, , , is the regression coefficient of the model, which quantifies the displacement caused by unit temperature, humidity and water level change, i.e. the quantitative correlation coefficient between environmental parameters and displacement change, is the constant term.

[0108] Next, using the constructed environmental displacement dataset, the model was fitted using methods such as least squares, and the specific coefficients in the compensation model for each point were solved. , , , .

[0109] Finally, in step S1034, the original displacement data is adjusted using the established compensation model. This process specifically includes:

[0110] The first step is to calculate the environmental displacement component. For each data point in the original displacement data, the corresponding environmental data (temperature, humidity, and water level changes) are used as independent variables and substituted into the compensation model for that point to calculate the environmental displacement component as the dependent variable. .

[0111] The second step is to generate the actual deformation information of the dam. This involves analyzing the total displacement corresponding to the original displacement data. In the middle, subtract the environmental displacement component calculated in the previous step. This yields the displacement that reflects the actual deformation of the structure: .

[0112] Displacement data after adjusting all measurement points The data is then reorganized according to its spatial location to generate the actual deformation information of the dam, which is then used as the second displacement data.

[0113] For example, the compensation model coefficient for a certain point on the waist of a dam is: =0.02mm / ℃, =0.005mm / %RH =0.1mm / m, constant term The value is 0. If, during a measurement, the temperature increases by 10°C, the humidity increases by 20%, and the water level rises by 2 meters, then the environmental displacement component can be calculated. Assume the total displacement at this point is measured by GNSS and total station. If the value is 2.5mm, then the actual displacement after compensation is... =2.5-0.5=2.0mm.

[0114] The overall scheme of S103 described above uses a total station driven by a dam surface model to accurately measure the original displacement data. Then, a compensation model based on multiple regression and environmental data are used for compensation, eliminating the interference of environmental factors on deformation monitoring. Finally, second displacement data that can accurately reflect the structural state of the dam is generated, providing a reliable data foundation for the subsequent construction of a spatial deformation field and anomaly early warning.

[0115] S104, obtain ground surface deformation data and settlement data, and generate comprehensive deformation data and settlement distribution data based on the ground surface deformation data, the second displacement data and the settlement data.

[0116] The ground surface deformation data is deformation amount data of the ground surface around the dam in a specific direction obtained by radar interferometry (InSAR) or the like.

[0117] The settlement data is absolute settlement amount data of the dam structure and its foundation body (i.e., the rock-soil body on which the dam is located and which interacts with the dam) in the vertical direction obtained by precise leveling or the like. The specific acquisition process includes: according to the dam geological survey data and structural design drawings, following the principle of controlling key positions and reflecting overall trends, arranging points. In the dam foundation area, monitoring sections are arranged at a set interval (such as 20 meters) along the dam axis direction to ensure that the point positions cover the full length of the dam foundation; in the dam abutment area, monitoring points are arranged at a certain distance (such as 15 meters) along the key geological boundaries according to the strike of the geological structure characteristics (such as faults, fissure zones, potential sliding surfaces); a high-precision electronic level and a matching indium strip code level are used to measure according to the specified leveling specification. When measuring, a stable reference point network is established, and the height difference between each monitoring point and the reference point is recorded each time; by comparing the height difference changes of the measurement data of different periods, the absolute vertical settlement of each monitoring point is calculated; finally, the vertical settlement data of all points is sorted to form a coordinated settlement data set for describing the coordinated deformation characteristics of the dam and the foundation body. For example, 20 points are arranged in the dam foundation, and 24 points are arranged in the dam abutment according to the geological conditions, and after precise leveling, accurate vertical settlement data of 44 points can be obtained.

[0118] The comprehensive deformation data is deformation data covering the dam and the surrounding area after fusing the second displacement data and the ground surface deformation data. The settlement distribution data is data showing the spatial distribution of points in the form of a settlement contour map by connecting points with equal settlement amounts in the dam body and the foundation body.

[0119] Optionally, step S104 can specifically include the following steps:

[0120] S1041, spatially superimpose the second displacement data and the deformation amount data of the plurality of points in the ground surface deformation data to generate comprehensive deformation data;

[0121] S1042, based on the comprehensive deformation data and the settlement data of the plurality of points in the dam and the foundation body, and using a scattered point-based surface reconstruction method, generate a continuous deformation surface;

[0122] The step S1042 specifically comprises: based on the comprehensive deformation data and the settlement data of multiple points in the dam and foundation body, combining into a scattered point data set; adopting a moving least square method to process the scattered point data set to generate the continuous deformation surface, the processing process comprising: for any target point on the continuous deformation surface, performing local surface fitting on the scattered point data set in the neighborhood of the target point, and determining the numerical value of the target point based on the result of the local surface fitting; wherein the neighborhood is determined by a search radius parameter, and in the dam foundation and dam shoulder area, the search radius parameter is adjusted according to the density and distribution characteristics of the monitoring points.

[0123] S1043, extracting settlement contour lines from the continuous deformation surface to generate settlement distribution data.

[0124] The step S1043 specifically comprises: extracting settlement contour lines from the continuous deformation surface according to a set settlement interval; after smoothing the extracted settlement contour lines, topologically connecting adjacent contour line segments to generate a settlement contour line map as the settlement distribution data.

[0125] In the embodiment of the application, first, the surface deformation data of the dam surrounding area is obtained by radar interferometric measurement technology; at the same time, by precise leveling, points are arranged at the key geological structure positions of the dam foundation and dam shoulder to obtain the settlement data of the dam and foundation body.

[0126] Secondly, by step S1041, the second displacement data (dam deformation data) obtained by step S103 and the obtained surface deformation data are spatially aligned through coordinate conversion, and are superimposed and fused by using a spatial interpolation algorithm (such as Kriging interpolation method) to generate comprehensive deformation data covering the dam structure and the surrounding area.

[0127] The process of superimposing and fusing by using the Kriging interpolation method to generate the comprehensive deformation data covering the dam structure and the surrounding area first needs to unify the coordinate systems of the second displacement data and the surface deformation data, and the manner for unifying the coordinate systems is a general manner in the art, which will not be described in detail herein. After unifying the coordinate systems, a “scattered point cloud” containing two groups of point data but still being discrete and uneven as a whole is obtained. The task of Kriging interpolation is to estimate the deformation value of any unknown point in the study area based on the deformation values of the known scattered points, so as to generate a continuous deformation field.

[0128] The core formula of Kriging interpolation is a weighted average:

[0129]

[0130] wherein, is the to-be-estimated point the deformation prediction value of the i-th known point. the deformation observation value of the i-th known point. is the weight assigned to the i-th known point. n is the number of known points within the neighborhood for estimation.

[0131] The essence of Kriging method lies in the way of determining the weight . Instead of simply using the inverse of distance, it is calculated through a tool called Semivariogram. Semivariogram describes the spatial correlation of data points, which calculates the half of the variance of attribute values between pairs of points at different distances h. By calculating the experimental semivariogram of all data points and fitting a theoretical model (such as spherical model, exponential model or Gaussian model), a mathematical function can be obtained to quantitatively describe the spatial structure characteristics of the entire region .

[0132] Based on this semivariogram model, Kriging interpolation obtains a set of weights with the smallest estimation variance by solving a system of linear equations. The form of this system of equations is as follows:

[0133]

[0134] where, is the semivariogram value between the known point and , is the semivariogram value between and , is the Lagrange multiplier introduced to meet the unbiased constraint.

[0135] By setting a regular grid for the study area and repeating the above Kriging interpolation calculation for each grid center point, the deformation prediction value of each grid point covering the entire dam and surrounding area can be obtained. This set of gridded prediction values ultimately constitutes the seamless and continuous integrated deformation data required.

[0136] Further, through step S1042, a continuous deformation surface is generated. The purpose is to construct a continuous and smooth mathematical surface based on a discrete, irregular scattered point data set, which can accurately represent the overall deformation state of the dam and its foundation. The process is as follows:

[0137] First, the integrated deformation data generated in the previous step and the obtained settlement data are merged in a unified coordinate system to form a scattered point data set containing all deformation and settlement information.

[0138] ​In the second step, the mobile least square method is used to process the scattered point data set. The core of the algorithm is that for any target point on the continuously deformed surface, it first determines a neighborhood range defined by the "search radius" parameter, and then in this neighborhood, it performs local polynomial surface fitting on the adjacent scattered data points by using the weighted least square method. Finally, the value of the target point (deformation or settlement) is determined based on the results of the local fitting surface. In this process, the key "search radius" parameter is dynamically adjusted according to the density and distribution characteristics of the monitoring points: in the key areas such as the dam foundation and abutment where the point density is high and the deformation gradient may be large, a smaller search radius is used to capture the fine deformation characteristics; while in the areas with sparse points, a larger search radius is used to ensure the continuity and smoothness of the surface.

[0139] Specifically, taking the dam safety monitoring of a large-scale water conservancy project as an example, to generate a continuously deformed surface of the dam body, the following operations are implemented: first, the dam body surface deformation data automatically collected based on the measuring robot and the foundation settlement data obtained by precise leveling are fused in a unified spatial coordinate system to form a three-dimensional scattered point data set that comprehensively represents the deformation of the dam body and foundation; second, the mobile least square method is used to reconstruct the surface of the data set, in the dam crest, gallery and abutment areas which are sensitive to deformation and have dense measuring points, a search radius of 15 meters is used to accurately capture the local warping and twisting deformation of these key parts caused by water pressure and temperature changes; in the far shore area behind the dam where the measuring points are relatively sparse, a search radius of 80 meters is used to construct a smooth transition surface using more extensive data points, effectively avoiding the fitting singular values caused by sparse data, and generating a high-precision, continuously smooth deformation surface model.

[0140] Finally, the settlement distribution data is generated from the generated continuously deformed surface through step S1043. The process is as follows: according to the accuracy requirements of dam safety monitoring, the settlement interval (such as 0.5 mm) is set, the points with settlement equal to the multiple of the interval are extracted from the continuously deformed surface, and the initial settlement contour is formed by connecting them. The extracted initial contour is smoothed by using Gaussian filtering algorithm and other algorithms to eliminate irregular fluctuations. Through topological inspection and connection, the adjacent contour segments after smoothing are correctly connected to finally form a complete settlement contour map that can clearly reflect the overall settlement distribution characteristics of the dam and foundation body, which is the settlement distribution data finally generated in this step.

[0141] The overall scheme of S104 above first generates comprehensive deformation data by fusing ground deformation data, second displacement data and settlement data, and then generates a continuously deformed surface by using an advanced surface reconstruction method, and finally outputs an intuitive and visual settlement distribution data (settlement contour map). This scheme realizes the integration of the deformation data of the dam and surrounding area, as well as the deformation data of the dam body and foundation body.

[0142] S105, extracting key features from the comprehensive deformation data and the settlement distribution data, processing the key features to identify abnormal patterns and predict deformation trends.

[0143] The key features are extracted from the comprehensive deformation data and the settlement distribution data, and include structured information of node attributes and topological connection relationships.

[0144] Optionally, step S105 can specifically include the following steps:

[0145] S1051, extracting key features from the comprehensive deformation data and the settlement distribution data; the key features include node attributes and topological connection relationships; the node attributes include spatial coordinates, settlement values and deformation values; and the topological connection relationships are used to represent the spatial distribution of the settlement distribution data.

[0146] S1052, inputting the key features into an autoencoder network, mapping the key features to a low-dimensional latent space through an encoder in the autoencoder network, and outputting reconstructed features from the low-dimensional latent space through a decoder in the autoencoder network.

[0147] The autoencoder algorithm is a kind of neural network for feature compression and reconstruction. The low-dimensional latent space is a low-dimensional data space output by the encoder. The reconstructed features are features restored by the decoder.

[0148] S1053, calculating reconstruction errors between the key features and the reconstructed features based on mean square errors.

[0149] The reconstruction error is the difference between the original key features and the reconstructed features, and is a key indicator for identifying abnormal patterns.

[0150] S1054, identifying the abnormal patterns based on the reconstruction errors and a dynamic threshold; the dynamic threshold is set based on historical normal data.

[0151] S1055, predicting the deformation trends based on historical deformation data.

[0152] In the embodiments of the present application, as Figure 2As shown, step S1051 first processes the settlement distribution data (settlement contour map) and the corresponding comprehensive deformation data to extract key features. Specifically, this process involves vectorizing the contour lines and extracting dense points from the vector data and comprehensive deformation data. For each point (node), multi-dimensional node attributes are recorded, including its spatial coordinates, the settlement value of its corresponding contour line, and the three-dimensional deformation obtained from the comprehensive deformation data. Simultaneously, a graph structure is used to represent the spatial distribution of the settlement distribution data. This is achieved by defining the connectivity (topological connection relationship) between nodes to reflect the continuous direction and spatial distribution pattern of the contour lines, and a global description can be performed using an adjacency matrix.

[0153] Next, in step S1052, the extracted key features are input into a pre-trained autoencoder network. This network compresses the high-dimensional key features through the encoder, mapping them to a low-dimensional latent space; subsequently, the decoder attempts to reconstruct the features with the same dimension as the original features from this low-dimensional latent space.

[0154] Then, in step S1053, the mean squared error (MSE) is used as the loss function to calculate the reconstruction error between the original key features and the reconstructed features. This error value quantifies the degree of information loss in the autoencoder when reconstructing the original features.

[0155] The specific calculation formula is as follows: ,in The first feature representing the original deep map The numerical value of the dimension. For the corresponding reconstructed feature representation, the first... Dimensional value, This represents the total dimension of the features. This error value measures the overall difference between the original features and the reconstructed features.

[0156] Next, through step S1054, abnormal patterns are identified based on the calculated reconstruction error. The specific method is as follows: First, reconstruction error data of the dam under historical normal operating conditions is collected, and its mean is calculated (…). ) and standard deviation ( ), and set a dynamic threshold accordingly (e.g. k is an adjustable coefficient, typically ranging from 1.5 to 3 (e.g., k=1.5). If the reconstruction error calculated in the current monitoring cycle exceeds this dynamic threshold, the dam's deformation is deemed to have entered an abnormal mode.

[0157] For example, take k = 1.5 as an example, the historical reconstruction error mean is 0.12, the standard deviation is 0.04, and the dynamic threshold is set to 0.12 + 1.5 x 0.04 = 0.18. If the reconstruction error calculated by S1053 is 0.2, since the reconstruction error 0.2 is greater than the dynamic threshold 0.18, it is determined that the deformation of the dam has an abnormal pattern.

[0158] Finally, through step S1055, the deformation trend is predicted based on the historical deformation data. This step is independent of abnormal identification and aims to predict the future deformation amount. The specific method is: collect the long-term historical deformation data sequence of the key points of the dam (such as monthly settlement), and use time series analysis model (such as ARIMA model) to model and extrapolate the sequence, thereby predicting the deformation trend of the dam in the future period.

[0159] The overall scheme of the above S105 extracts key features from comprehensive deformation data and settlement distribution data, and calculates reconstruction error combined with autoencoder algorithm, realizes intelligent identification of abnormal pattern of dam deformation. At the same time, the time series analysis method is combined to predict the deformation trend. The scheme converts complex visual data into quantifiable features, improving the intelligent level of dam deformation monitoring and the accuracy of abnormal identification.

[0160] S106, generating warning information based on the abnormal pattern and the deformation trend.

[0161] Among them, the warning information is a prompt content containing the current warning level and the abnormal deformation position, used to inform the staff of the safety risk situation of the dam.

[0162] In the embodiments of the present application, first, a plurality of warning levels are set, which are usually divided into blue, yellow, orange and red four levels, and each warning level corresponds to different deformation threshold and abnormal pattern type. For example, blue warning corresponds to smaller deformation threshold and slight abnormal pattern, while red warning corresponds to the largest deformation threshold and serious abnormal pattern.

[0163] Secondly, the deformation trend (such as the dam settlement in the next 3 months) output by step S105 is compared with the deformation threshold of each warning level set, and the abnormal pattern identified is combined to comprehensively judge the safety state of the current dam.

[0164] Finally, based on the above comprehensive judgment result, the current warning level is determined, and the warning information is generated. For example, if the judgment result reaches the standard of yellow warning, the warning information containing the warning level (yellow warning) and the corresponding position of the abnormal pattern (such as the left area of the dam waist) is generated, so that the staff can know the specific risk situation.

[0165] In practical applications, in the monitoring of the A-type concrete dam, four early warning levels of blue, yellow, orange and red and their corresponding deformation thresholds and abnormal patterns are set. The deformation trend output by step S105 is "the settlement of the right side area of the dam abutment in the next month is 0.8mm", and the "moderate abnormal pattern" is identified. The result meets the starting condition of yellow early warning, so the current early warning level is determined to be yellow early warning, and the early warning information is generated: "yellow early warning: the estimated settlement of the right side area of the dam abutment in the next month reaches 0.8mm, and there is a moderate deformation anomaly".

[0166] The overall scheme of S106 described above can clearly and intuitively reflect the safety risk level and specific abnormal position of the dam by setting a hierarchical early warning mechanism, generating targeted early warning information in combination with the deformation trend and the abnormal pattern, providing clear warning basis for dam safety management personnel, facilitating them to take corresponding targeted measures such as investigation and maintenance in time, and effectively preventing the occurrence of dam safety accidents.

[0167] Figure 3 A specific implementation structure diagram of a concrete dam high-precision deformation monitoring system fusing GNSS and external environmental parameters provided by the embodiments of the present application is shown in Figure 3 The system can include:

[0168] The acquisition module 31 is configured to acquire first displacement data and environmental data. The first displacement data is displacement data of a dam monitoring point, and the environmental data includes temperature data, humidity data and water level data.

[0169] The generation module 32 is configured to acquire photogrammetry data, generate a dam surface model based on the photogrammetry data and the first displacement data, perform measurement based on the dam surface model to obtain original displacement data, generate second displacement data based on the original displacement data and the environmental data and using a compensation model, and acquire ground surface deformation data and settlement data, generate comprehensive deformation data and settlement distribution data based on the ground surface deformation data, the second displacement data and the settlement data.

[0170] The identification module 33 is configured to extract key features from the comprehensive deformation data and the settlement distribution data, process the key features to identify abnormal patterns and predict deformation trends.

[0171] The generation module 32 is further configured to generate early warning information based on the abnormal patterns and the deformation trends.

[0172] The concrete dam high-precision deformation monitoring system fusing GNSS and external environment parameters according to the embodiments of the present application is used to implement the aforementioned concrete dam high-precision deformation monitoring method fusing GNSS and external environment parameters, and therefore the specific embodiments of the concrete dam high-precision deformation monitoring system fusing GNSS and external environment parameters can refer to the embodiments of the concrete dam high-precision deformation monitoring method fusing GNSS and external environment parameters in the foregoing description, and will not be repeated here.

[0173] The present application further provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the aforementioned concrete dam high-precision deformation monitoring methods fusing GNSS and external environment parameters.

[0174] The present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of any of the aforementioned concrete dam high-precision deformation monitoring methods fusing GNSS and external environment parameters.

[0175] In an exemplary embodiment, the aforementioned computer readable storage medium can include, but is not limited to, a U disk, a read-only memory, a random access memory, a mobile hard disk, a magnetic disk or an optical disk, and various media capable of storing computer programs.

[0176] The embodiments of the present application further provide a computer program product, which comprises a computer program, and the computer program is executed by a processor to implement the steps in any of the aforementioned concrete dam high-precision deformation monitoring methods fusing GNSS and external environment parameters.

[0177] The skilled person can further realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been described in general terms in the foregoing description. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0178] The above describes in detail the method and system for high-precision deformation monitoring of a concrete dam by fusing GNSS and external environmental parameters. The principles and implementation modes of the present application are described by using specific examples, and the above description of the examples is only used to help understand the method and its core idea. It should be pointed out that, for those skilled in the art, without departing from the principles of the present application, the present application can be improved and modified in several ways, and these improvements and modifications also fall within the protection scope of the present application.

Claims

1. A method for high-precision deformation monitoring of a concrete dam by fusing GNSS and external environmental parameters, characterized in that, The method comprises: obtaining first displacement data and environmental data, the first displacement data being displacement data of a dam monitoring point collected by a GNSS receiver, and the environmental data including temperature data, humidity data and water level data; obtaining photogrammetry data, generating a dam surface model based on the photogrammetry data and the first displacement data; based on the dam surface model, measuring to obtain original displacement data, and based on the original displacement data and the environmental data, generating second displacement data using a compensation model; the second displacement data is the deformation data of the dam; obtaining ground surface deformation data and settlement data, generating comprehensive deformation data and settlement distribution data based on the ground surface deformation data, the second displacement data and the settlement data; extracting key features from the comprehensive deformation data and the settlement distribution data, processing the key features to identify abnormal patterns and predict deformation trends; generating early warning information based on the abnormal patterns and the deformation trends; based on the dam surface model, measuring to obtain original displacement data, and based on the original displacement data and the environmental data, generating second displacement data using a compensation model, comprising: based on the dam surface model, determining the measurement point positions of the total station; driving the total station to perform laser measurement on the measurement point positions, obtaining a three-dimensional coordinate sequence of each measurement point position through multiple laser measurements to generate a displacement change sequence, and combining the displacement change sequence into original displacement data; establishing a compensation model of environmental parameters and displacement changes, the compensation model being established based on a multiple regression model; using the compensation model and based on the environmental data, adjusting the original displacement data to generate real deformation information of the dam, and taking the real deformation information of the dam as the second displacement data; based on the ground surface deformation data, the second displacement data and the settlement data, generating comprehensive deformation data and settlement distribution data, comprising: spatially superimposing the second displacement data and the deformation amount data of multiple point positions in the ground surface deformation data to generate comprehensive deformation data; based on the comprehensive deformation data and the settlement data of multiple point positions in the dam and the foundation body, and using a scattered point-based surface reconstruction method, generating a continuous deformation surface; extracting settlement contour lines from the continuous deformation surface to generate settlement distribution data; from the comprehensive deformation data and the settlement distribution data, extracting key features, processing the key features to identify abnormal patterns and predict deformation trends, comprising: extracting key features from the comprehensive deformation data and the settlement distribution data; wherein the key features include node attributes and topological connection relationships, the node attributes include spatial coordinates, settlement values and deformation values, and the topological connection relationships are used to represent the spatial distribution of the settlement distribution data; inputting the key features into an autoencoder network, mapping the key features to a low-dimensional latent space through an encoder in the autoencoder network, and outputting reconstructed features from the low-dimensional latent space through a decoder in the autoencoder network; calculating a reconstruction error between the key feature and the reconstructed feature based on a mean square error; identifying the abnormal pattern based on the reconstruction error and a dynamic threshold; the dynamic threshold is set based on historical normal data; predicting the deformation trend based on historical deformation data.

2. The method of claim 1, wherein, generating a continuous deformation surface based on the comprehensive deformation data and settlement data of multiple points in the dam and foundation body, including: combining the comprehensive deformation data and settlement data of multiple points in the dam and foundation body into a scattered point dataset; processing the scattered point dataset to generate the continuous deformation surface using a moving least squares method, including: for any target point on the continuous deformation surface, performing local surface fitting on the scattered point dataset in the neighborhood of the target point, and determining the value of the target point based on the result of the local surface fitting; wherein the neighborhood is determined by a search radius parameter, and in the dam foundation and abutment area, the search radius parameter is adjusted according to the density and distribution characteristics of the monitoring points.

3. The method of claim 1, wherein, extracting settlement contour lines from the continuous deformation surface to generate settlement distribution data, including: extracting settlement contour lines from the continuous deformation surface according to a set settlement interval; topologically connecting adjacent contour line segments after smoothing the extracted settlement contour lines to generate a settlement contour line map as settlement distribution data.

4. The method of claim 1, wherein, using the compensation model and adjusting the original displacement data based on the environmental data to generate real deformation information of the dam, including: using the compensation model to calculate the environmental displacement component with temperature, humidity and water level change as independent variables and the environmental displacement component as dependent variable; subtracting the environmental displacement component from the original displacement data to generate real deformation information of the dam.

5. The method of claim 1, wherein, acquiring photogrammetry data, generating a dam surface model based on the photogrammetry data and the first displacement data, including: acquiring image data sets by multi-angle image acquisition of the dam surface by a camera carried by a drone; extracting feature points from the image data sets, calculating initial three-dimensional point clouds according to the matching relationship of the feature points; using the first displacement data as control points to correct the position of the initial three-dimensional point clouds to generate corrected three-dimensional point clouds; generating dense point clouds based on the corrected three-dimensional point clouds, and constructing a dam surface model based on the dense point clouds.

6. A high-precision deformation monitoring system for concrete dams that combines GNSS and external environmental parameters, characterized in that, including: an acquisition module for acquiring first displacement data and environmental data, the first displacement data being displacement data of dam monitoring points, and the environmental data including temperature data, humidity data and water level data; a generation module for acquiring photogrammetry data, generating a dam surface model based on the photogrammetry data and the first displacement data; measuring based on the dam surface model to obtain original displacement data, and generating second displacement data based on the original displacement data and the environmental data using a compensation model; The second displacement data is deformation data of the dam; ground surface deformation data and settlement data are obtained, and based on the ground surface deformation data, the second displacement data and the settlement data, comprehensive deformation data and settlement distribution data are generated; The recognition module is configured to extract key features from the comprehensive deformation data and the settlement distribution data, process the key features, identify an abnormal pattern and predict a deformation trend; The generation module is further configured to generate early warning information based on the abnormal pattern and the deformation trend; The generation module is further configured to generate the second displacement data based on the original displacement data and the environmental data, and using a compensation model, including: Based on the dam surface model, determining the measurement point positions of the total station; Driving the total station to perform laser measurement on the measurement point positions, obtaining a three-dimensional coordinate sequence of each measurement point position through multiple laser measurements to generate a displacement change sequence, and combining the displacement change sequence into the original displacement data; Establishing an environmental parameter and displacement change compensation model, the compensation model being established based on a multiple regression model; Using the compensation model and based on the environmental data, adjusting the original displacement data to generate real deformation information of the dam, and taking the real deformation information of the dam as the second displacement data; The generation module is further configured to generate the second displacement data based on the original displacement data and the environmental data, and using a compensation model, including: Spatially superimposing the second displacement data and the deformation amount data of multiple point positions in the ground surface deformation data to generate comprehensive deformation data; Based on the comprehensive deformation data and the settlement data of multiple point positions in the dam and the foundation body, and using a scattered point-based surface reconstruction method, a continuous deformation surface is generated; Extracting a settlement contour from the continuous deformation surface to generate settlement distribution data; The recognition module is further configured to extract key features from the comprehensive deformation data and the settlement distribution data, process the key features, identify an abnormal pattern and predict a deformation trend, including: Extracting key features from the comprehensive deformation data and the settlement distribution data; wherein the key features include node attributes and topological connection relationships, the node attributes include spatial coordinates, settlement values and deformation values, and the topological connection relationships are used to represent the spatial distribution of the settlement distribution data; Inputting the key features into a self-encoder network, mapping the key features to a low-dimensional latent space through an encoder in the self-encoder network, and outputting reconstructed features from the low-dimensional latent space through a decoder in the self-encoder network; Based on the reconstruction error and a dynamic threshold, identifying the abnormal pattern; the dynamic threshold is set based on historical normal data; Based on historical deformation data, predicting the deformation trend. The memory is configured to store a computer program.

7. An electronic device, comprising: ​ ​ A processor for implementing the steps of a method for high-precision deformation monitoring of a concrete dam by fusing GNSS and external environmental parameters according to any one of claims 1 to 5 when executing the computer program.

Images